Keep missing Swift timing distinct from absent invocation

docs/src/components/LatencyTier.astro

@@ -1238,7 +1238,7 @@ interface Props {
 
 <p>
   Of the case results in this bound latency corpus, <strong>{model.totalTimed}</strong>
-  invoked an analyzer and carry a timing sidecar.{' '}
+  carry a retained timing sidecar.{' '}
   <strong>{model.totalUntimed}</strong> do not, and every one of them is
   accounted for:
 </p>
@@ -1268,8 +1268,10 @@ interface Props {
   is nothing to time. Those cases are <strong>absent from every distribution
   on this page, not entered as zero</strong> — entering them as zero would
   make the adapters that decline the most look the fastest, which is exactly
-  backwards. Every case that did invoke an analyzer is timed; the timed and
-  untimed columns account for the whole bound population with no remainder.
+  backwards. A missing sidecar does not establish that analysis never ran;
+  Swift reference results can retain an outcome without this timing format.
+  Missing timings stay outside distributions. The timed and untimed columns
+  account for the whole bound population with no remainder.
 </p>
 
 </>)}

docs/src/data/latency.ts

@@ -101,9 +101,9 @@ export interface LatencyModel {
   environments: EnvironmentStamp[];
   adapters: LatencyAdapter[];
   totalTimed: number;
-  /** Bound cases with no timing, because they never invoked the analyzer. */
+  /** Bound cases without a retained timing sidecar. */
   totalUntimed: number;
-  /** Outcomes those untimed cases carry — expected to be `unsupported` only. */
+  /** Outcomes those untimed cases carry, including incomplete coverage. */
   untimedOutcomes: { outcome: string; cases: number }[];
 }
 
@@ -211,7 +211,7 @@ interface CaseTiming {
 
 /**
  * One case's timing sidecar, or `null` when the freeze does not bind the case
- * or the case never invoked an analyzer (and so has nothing to time).
+ * or no sidecar was retained. Missing timing does not prove no invocation.
  */
 function readTiming(
   evidence: ArchivedLatencyEvidence,